I originally thought this was an issue with pandas ( pandas-dev/pandas#12029 ), but I can reproduce it by using matplotlib alone.

After changing the style to "fivethirtyeight" (with pyplot.style.use("fivethirtyeight") ) and changing the default colormap (via rcParams), both bar plots and boxplots still use the default colors:

plt.plot([0, 1, 2, 3], [5, 6, 7, 8])
plt.plot([0, 1, 2, 3], [6, 3, 7, 12], alpha=0.5)

plt.bar([0, 1, 2, 3], [5, 6, 7, 8])
plt.bar([0, 1, 2, 3], [6, 3, 7, 12], alpha=0.5)

plt.boxplot([6, 3, 7, 12, 20, 25, 33])

The code I used to customize the plotting style can be found on the pandas issue (like I said, it's mostly just setting the style to fivethirtyeight then changing the color cycle)